Data Sharing Guidance for Public Transit Agencies—Now and in the Future (2020)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

CHAPTER 4 Models for Sharing Public Transit Data With the wide variety of data types that transit agencies collect, and the often resource and time-intensive processes required to prepare data for sharing, transit agencies are faced with decisions about which data to share with which audiences. This chapter describes different models for transit data sharing, identified both from the review of literature and web sources and from the transit agency interviews. Data sharing often involves public–private partnerships, which have mixed assessments in the literature. Public–private partnerships have been used to decrease costs and increase efficiency (Lin and Mele 2012). Generally, they occur when the public entity has a capacity gap to fill. In the case of data sharing, either they have data they cannot interpret, or they want to use a private company’s data to enhance decisionmaking (Mackintosh 1992). Broadly, models for sharing data can be classified as public or private. Public models, often referred to as open data, make data available to everyone, typically by publishing it online. In private models, data is shared with an individual, institution, or group of individuals or institutions. Much of the time, this data is shared under a data agreement, often including clauses about nondisclosure, preventing the data from being shared more widely. Within each of these classes of models there is considerable variation, as described in the sections that follow. Advantages and disadvantages of public and private data sharing are summarized in Figure 5 and elaborated on in the following sections. 4.1  Public Data Sharing Public data sharing is prevalent among transit agencies. In a 2015 study that surveyed 67 transit agencies, 83% provided open data (Schweiger 2015). All the transit agency interview- ees stated they publish data on their websites. Sharing data openly promotes transparency and can spur innovative use of their data. Additionally, public data sharing was touted by transit agency interviewees for its efficiency. Transit agencies receive large volumes of public records and other data requests, and publishing the data online allows transit agency staff to quickly point requesters to the online portal rather than have to fulfill requests individually. Data users may also download data directly from transit agency websites without interacting with the transit agency at all. Although there is an upfront cost to putting data online and providing necessary documentation, transit agencies believe putting frequently requested, nonconfidential data online saves them staff time in the long term. According to the literature, open data programs increase the public’s perception of account- ability. The assumption is, if the government is doing something wrong, it will show in the data (Brauneis and Goodman 2017). In addition, open data can lead to “citizen-sourcing” or utilizing the collective knowledge of residents to analyze and interpret data being released (Kassen 2013). 36 

Models for Sharing Public Transit Data    37   Figure 5.   Advantages and disadvantages of public and private data sharing. By building data portals that respond to citizen needs and requests, cities can build cycles of trust and stewardship of their data (Dawes 2010; Homstad 2018). On the other hand, posting data pub- licly generally means that public transit agencies cede control over how the data is used. Lack of knowledge about the data available and lack of public capacity to utilize specific data types can also limit its impacts (Shelton et al. 2014). One innovative solution to this challenge was a data competition that one interviewed transit agency hosted. Not only did this bring attention to the transit agency’s open data, but it encouraged participants to compete to answer questions posed by the transit agency. Several transit agencies are also considering different methods to ensure the route, schedule, and vehicle location data they publish are used to best serve their customers (see Section 4.3). Transit Data Types That Are Shared Online The data that transit agencies publish online includes route and schedule information, system alerts, and the real-time location of transit vehicles. In most cases, route and schedule informa- tion is published in the standard GTFS format, and vehicle location data follows the GTFS-RT standard. According to the 2015 study, the most common examples of open data among transit agencies are route and schedule information and vehicle location feeds (Schweiger 2015). In addition to route, schedule, and vehicle location data, many transit agency interviewees indicated their agencies publish information on performance indicators, including route or line level ridership, passenger counts at bus stops and train stations, on-time performance, and reliability indicators. Transit agencies also provide summaries of survey data, including travel surveys and customer satisfaction surveys. Finally, at least one transit agency interviewee indicated that their agency publishes financial data. Online Sharing Formats: Reports, Repositories, Dashboards, and Application Programming Interfaces Transit agencies share data publicly in a variety of forms. All the transit agency interviewees indicated their agencies have data and reports that can be downloaded from their websites. In addition, two have interactive dashboards that allow users to interact with the data in a con- trolled way. Most of the transit agencies share route, schedule, and vehicle location data using an API, which is essentially a set of methods for retrieving data that makes it easy for developers to use the data. These formats have advantages and disadvantages, as shown in Figure 6. Although static reports are easy for all audiences to understand, they do not allow researchers and innovators to manipulate the data, which can limit new insights that could be drawn from the underlying 

38   Data Sharing Guidance for Public Transit Agencies—Now and in the Future Figure 6.   Methods for sharing data online. data. On the other hand, static reports protect against data misuse, because the analysis is performed by transit agency staff. Interactive dashboards typically also limit the chances of misinterpretation of data, because they allow for only controlled data manipulation. For example, one transit agency interviewee indicted their agency has a dashboard that allows users to look at service reliability and ridership for a specific route, date, and period of the day. Backend calculations are programmed by the transit agency, preventing incorrect analysis of the underlying data. Dashboards can be a conve- nient way for people with a wide range of technical abilities to interact with transit agency data. Of course, if the underlying data is not provided, dashboards do not promote new analysis. For example, although the dashboard described allows the user to retrieve reliability information, defined as the on-time percentage for a route (for low-frequency service), the underlying data would enable a researcher to answer more detailed questions about the extent and patterns of schedule deviation. In addition, developing an interactive dashboard requires significant effort on the part of the transit agency. In some cases, third parties have produced public dashboards based on open data. For example, the Bus Turnaround Coalition developed a dashboard that reports on the performance of New York City’s bus routes. Most transit agency interviewees noted their agencies have a developer website designed for use by software developers. These sites house the transit agency’s API. Developers use these interfaces to access data that is then provided to customers in travel planning and real-time information apps. Some of the transit agencies require the users of this data to register to access an API key. At least two of the transit agency interviewees noted that this model enables their agency to cut off users who overburden the system with too many data requests. Travel plan- ning and real-time information apps have become a key source of transit customer informa- tion. As such, transit agencies are reconsidering the best way to leverage their route, schedule, and vehicle location data to provide customers with information. Section 4.3 describes this debate in detail. Many transit agencies use a combination of the mechanisms in Figure 6 to share data and information publicly. For example, the MBTA’s interactive performance dashboard allows users to select specific lines, dates, and periods when viewing reliability and ridership data. In addition, the underlying data is available for download. The transit agency also has a developer API and publishes reports that share transit agency insights and analysis of their data. 

Models for Sharing Public Transit Data    39   Developing a Dashboard for Multiple Audiences Providing data that is relevant to citizens and providing it in forms that are usable and responsive is a challenge for public agencies (Abella et al. 2017). It is possible to have a highly interactive dashboard that provides data in multiple formats. One such example comes from Pecan Street, a non-profit research and development organization that hosts electricity network data from public, academic, and commercial sources. Pecan Street developed Dataport as a platform to manage users, permissions, and data. Pecan Street’s greatest challenge to date is incorporating new data resources without increasing the complexity of the platform, which supports online analyses. It began with energy data, and grew to include time-stamped electricity, water, gas, solar, weather, and transportation data. Pecan Street recognizes it has two types of users: “power” users and non- power users. Power users can directly query the database and join any data sets, thus using and manipulating any data set for insights. Non-power users can query and download data in Excel sheets. Cross-Agency Data Sharing Platforms Multiple public transit agencies share data with an organization that shares the data using a centralized platform. Such platforms typically require a standardized data format and may therefore require additional effort from transit agencies. Conversely, such models can allow transit agencies to share costs of processing, storing, and documenting data as well as addressing any legal implications of data sharing, potentially reducing transit agency effort. FTA’s NTD is an example of a cross-agency data repository. Recipients of FTA grants are required to submit transit system, ridership, and financial data to the database. The standardized format makes it highly accessible to researchers, who produce studies that can benefit transit agencies. The National Association of City Transportation Officials (NACTO), which includes both cities and transit agencies, launched the SharedStreets initiative in 2018 (National Association of City Transportation Officials 2018). SharedStreets is an organization that provides open-source software tools and digital infrastructure that allows public entities and private companies to manage and share data about their physical infrastructure and vehicle activity. SharedStreets highlights four core functions of its collaborative platform: (1) to standardize data on physical infrastructure and vehicle activity, (2) to build open-source tools to use data, (3) to anonymize sensitive data on individuals, and (4) to establish a foundation of collaboration and trust (National Association of City Transportation Officials 2018). These functions illustrate the potential benefit of cross-agency platforms to transit agencies. Efforts to standardize and anonymize data can be shared across agencies, and open-source tools that operate on standardized data can provide value back to these transit agencies. These initiatives also represent a way for transit agencies to access private data. Uber, Lyft, and Ford Motor Company are all involved in the SharedStreets initiative. Chapter 5 describes transit agency access to external data sources in more detail. Many state and local governments have created robust online repositories of their data. Some cities are automating and centralizing their data upload process. These repositories not only enable public access to data, but they facilitate the use of data across agencies (e.g., providing cities with access to transit agency data and vice versa). 

40   Data Sharing Guidance for Public Transit Agencies—Now and in the Future Terms of Use for Public Data The inclusion of terms of use or legal notices with open data provided online varies among transit agencies. For example, the performance data that powers the MBTA performance dash- board can be downloaded without agreeing to any terms of use. Houston METRO, however, includes a legal notice with its data downloads, indicating that METRO retains ownership of the data and that the data is provided “as is.” TransLink has similar terms of use for their Open API data (https://developer.translink.ca/Home/TermsOfUse). Section 2.4 provides guidance on the inclusion of terms of use, including model terms to protect transit agencies providing open data. 4.2 Private Data Sharing Some data is more sensitive but still sharable under the right conditions. Other data types are not commonly requested and therefore have not been published on the transit agency’s website but can be shared when requested. To fulfill these needs, all the transit agency inter- viewees indicated their agencies share data directly with partners and data requesters. This category of data sharing includes data sharing with partners, such as research institutions, municipalities, and private sector contractors or real estate developers, and fulfillment of public records requests. When sharing sensitive data with partners, transit agencies may require nondisclosure agreements and training prior to releasing the data. Research Partnerships Most transit agency interviewees described the importance of sharing data with researchers. Five interviewees indicated their agencies have strong partnerships with a specific university or research institute. In at least two cases, the transit agency pays the university to complete research that aligns with the transit agency’s needs. In other cases, transit agencies do not pay the university, but nonetheless have a long-standing collaborative relationship that allows them to shape their partners’ research agendas. In some cases, transit agencies are willing to share sensi- tive data, including individual customer records, with researchers in these universities, after the researchers are trained and have signed nondisclosure agreements. Researchers are often able to spend time on questions that transit agencies are unable to, providing significant benefits to the transit agency. Long-standing partnerships allow the transit agency to have a standing data agreement and an established trust in the researchers. Other Data Sharing Relationships At least two of the transit agency interviewees indicated that their agencies shared data with local municipalities. In addition, several of the transit agencies provide data to real estate devel- opers when requested. One transit agency interviewee specifically mentioned advertisers as a recipient of data. Another private data sharing model is followed by the American Bus Benchmarking Group. Established in April 2011, the group consists of midsized bus organizations in the United States. Members share performance data that can be accessed confidentially by members of the group. The objective is to establish benchmarks that can help members understand their transit agency’s performance and identify best practices to improve performance (American Bus Benchmarking Group 2019). Terms of Use for Private Data Sharing As noted, sensitive data shared with a partner is typically accompanied by a nondisclosure agreement, according to the transit agency interviewees. In addition, at least two transit agencies 

Models for Sharing Public Transit Data    41   require any publications from their academic partners to be reviewed by the transit agency prior to publication. On the other hand, it is common to provide nonsensitive data shared through public records requests without any provisions for use. At least two transit agency interviewees noted their agencies do not attach any provisions when they fulfill nonsensitive data requests, which may include aggregate statistics on ridership or boardings and transit system-level information. 4.3 Examples of Data Sharing Models for Customer Information Route, schedule, and vehicle location data are among the most commonly shared types of transit data. According to information given in TCRP Synthesis 115, most transit agencies share this data and do so free of cost (Schweiger 2015). This data is typically shared in standardized GTFS and GTFS-RT formats, and this information has a clear value to customers planning transit trips and finding out when transit vehicles will arrive. Across the United States and abroad, private app developers have created apps that use GTFS and GTFS-RT feeds to provide information to customers, and this is one of the most prominent examples of transit agencies sharing data. In London alone, there are 600 apps powered by public transit open data feeds, which are used by 42% of Londoners (Deloitte 2017). From TCRP Synthesis 115, approximately 40% of respondents to an APTA survey have developers using their open data. For large transit agencies, 68% reported that developers “Riders interact with these apps multiple use their data (Schweiger 2015). times daily, making open data the most important customer communication These apps have become a key component of how customers interact channel agencies offer to the public.” with transit systems. In many cases, customers are much more likely (TransitCenter 2018) to receive information about transit services from private apps than directly from the transit agency. A review of different models for sharing data with app developers shows how some transit agencies are leveraging this data to exert more control over how apps serve their customers. In addition, transit agencies are developing models in which they receive data collected by transportation apps using a variety of mechanisms to facilitate this data transfer. Figure 7 shows four different models for using route, schedule, and vehicle location data to provide customer information as reported in the transit agency interviews. Transit agencies can control the customer information received through an app in a variety of ways. Many transit agency interviewees indicated their agencies simply publish route, schedule, and vehicle loca- tion data and allow app developers to use it in their apps (dubbed an app-neutral approach). A relatively low-effort option that provides the transit agency with some control is to endorse an existing app. A more resource-intensive option, which one transit agency is in the process of completing, is to commission an app. Finally, for ultimate control, some transit agencies, includ- ing two of those interviewed, develop apps in house. This requires staff with specific technical skills to develop and maintain the app. It is important to note that, in the cases observed and documented, those transit agencies that endorse, commission, or develop an app continue to provide the data openly and allow other apps to use the data. App-Neutral Approach Many transit agencies publish GTFS and GTFS-RT feeds using a developer API, which in turn is used by third-party apps. Some transit agencies provide a list of vetted apps on their websites. Even with this safeguard, this hands-off data sharing model relies on developers to provide the 

42   Data Sharing Guidance for Public Transit Agencies—Now and in the Future Figure 7.   Mechanisms for customer information app development. best information and user experience to customers and customers to find the best app through the app review process or word of mouth. There are drawbacks to this approach, many of which were highlighted by transit agency interviews conducted. First, some apps may not provide accurate information. Even if all apps use the same data feeds, the algorithms they use to suggest routes and predict vehicle arrivals vary. One transit agency interviewee commented that one commonly used transportation app in their service area often provided inaccurate predictions, likely due to the algorithm used. This is a problem, because transit agencies want to ensure that customers access the most reliable information available. Transit agencies are also concerned with the context under which information is presented. Third-party apps exclusively control the content and presentation of information and may rely on advertising revenue from other modes of transportation that are included in the app. Many apps, including Transit App and Google Maps, present information about the cost and travel time of Uber and Lyft on the same screen as transit information. Informed by customer research, at least one transit agency interviewee expressed concerns that this presentation of informa- tion may encourage users to choose TNCs over public transit. On the other hand, some transit agencies see TNCs as a potential complement and have worked to have transit information displayed within the Uber app. In Denver, not only are public transit options listed alongside Uber’s offering in the Uber app, but customers can purchase tickets for these services in the app as well (Conger 2019). A hands-off approach to sharing GTFS and GTFS-RT data also means that data on transit cus- tomers who use transportation apps is collected by the app developers rather than by the transit agency. The UITP argues that this can be strategically risky for transit agencies, because they are missing out on information about their customers. Instead, this information is accumulating to private developers. There is at least one example in which a private app developer, Citymapper, piloted bus service; however, in this example, the pilot was in cooperation with TfL, a public transit agency. In the “endorse an app” example that follows, the transit agency was able to negotiate access to a third party’s app data in return for endorsing that app. 

Models for Sharing Public Transit Data    43   Endorse an App in Return for Data In Boston, Transit App, which the MBTA officially endorsed for real-time transit informa- tion and travel planning, provides the MBTA with data on app users’ searches and locations (Enwemeka 2016). The MBTA made this agreement after releasing a nonfunded request for proposal (RFP) asking app developers to submit bids competing for endorsement by the tran- sit agency. The transit agency evaluated bids and selected the best app to recommend to its users. The RFP also required that the selected app share data with the transit agency, enabling the transit agency to draw insights from app data. Compared with an in-house or commissioned app, this model is much less expensive. How- ever, as in the other cases, there is no guarantee that customers will use the endorsed app. In addition, in this model, the transit agency has less control over the presentation of information in the app. Rather than specifying these components explicitly, they select the best of the avail- able options. Commission App Development Although most transportation apps are developed by private app developers without input or funding from transit agencies, transit agencies may opt to pay for some or all components of app development, giving them control over app specifications. Commissioned apps may be branded by the third-party developer or by the transit agency itself. One transit agency interviewee indicated that their agency had released an RFP seeking a vendor to provide a trip planner to integrate into the agency’s existing mobile payment app. Not only will the transit agency have control over the way information is presented, but they anticipate being able to provide more custom information such as detours, real-time alerts, and service change notifications. The RFP also specifies performance requirements, including a minimum level of prediction accuracy and a limit on the “ghost” bus and train incidence rate. The transit agency will own all data produced by the app. The transit agency anticipates that the data will improve the transit agency’s understanding of operations, predictions accuracy, customer preferences, and travel patterns. For example, this data could provide insights into customer origins and destinations, travel choices, and latent demand as well as customer responsiveness to changes in routes, frequency, service quality, and reliability. To protect user privacy, the vendor must abide by stringent General Data Protection Regulation (GDPR) standards now employed in the European Union. A major drawback of this approach is the cost. In addition, the transit agency cannot guaran- tee that customers will use the app, because other apps will still be available. Also, one argument against a custom app, whether it is developed in house or commissioned, is that it works against the trend of universal transit planning tools. Apps like Google Maps and Transit App that are available across a large number of transit markets standardize the transit experience for visitors who ride transit in different cities. Standardization of apps may make transit easier to use and actually encourage transit use. Develop an App in House Two transit agency interviewees indicated their agencies have their own app, developed in house. This model gives the transit agency maximum control over the contents of the app and the data extracted from it. As long as the transit agency continues to put resources into maintain- ing and updating the app, they have flexibility to adapt the app over time, as data and customer needs change. 

44   Data Sharing Guidance for Public Transit Agencies—Now and in the Future Such a model is generally only feasible for large transit agencies with sufficient information technology (IT) and technical staff to devote to the effort. However, even some of the transit agency interviewees from larger agencies indicated their agencies opted out of developing their own app, because of the specialized and fast-changing nature of app development. Transit agencies that either commission an app or develop it in house make decisions about the inclusion of other modes. On the one hand, some see a benefit to excluding potentially com- peting modes such as TNCs. On the other hand, several transit agency interviewees reported that they believe it is important to include other modes to facilitate multimodal journeys. In regions with multiple transit providers, integration of data from other agencies into a commissioned or in-house app is critical.